Verify AI Output

Most AI systems produce answers you cannot verify. AIEP changes that.

With AIEP, every AI response carries its own proof: the sources it used, the hash of those sources, and a validation result you can check independently.

The verification problem

When an AI answers a question, you typically receive:

The answer text
Possibly some source names
No way to confirm the sources are unchanged
No audit trail

If the answer is wrong, or if the source has been updated, you have no way to detect it.

How AIEP verification works

AIEP binds each AI response to an evidence rail — an ordered list of source artefacts, each hashed and timestamped.

Input question
    ↓
AI generates response
    ↓
Evidence artefacts retrieved and hashed (SHA-256)
    ↓
Response bound to evidence rail
    ↓
Schema validation run
    ↓
Validation result: PASSED / FAILED

Every step is recorded. The result can be replayed.

What you can verify

Field	What it proves
`source_url`	Where the evidence came from
`content_hash`	The exact content at retrieval time
`retrieved_at`	When the artefact was fetched
`validation_status`	Whether the artefact passed schema checks
`confidence_tier`	Evidence quality classification

Try it

Verification Playground → — run a sample verification end-to-end

See an example verified response →

View the response JSON example →

View on GitHub →